Proposal of an improved motion estimation module for SVC
Ronaldo Hüsemann, Ricardo Kintschner, Valter Roesler, José Valdeni de Lima, Altamiro Susin · 2010
In October 2007 ITU-T and MPEG organizations published the Scalable Video Coding (SVC) specification as an addendum to the H.264 standard. Basically SVC enables H.264 video encoders to support temporal, spatial and SNR scalability. This new functionalities lead to more flexibility on bandwidth management and robustness against packet losses. It is therefore more suitable for heterogeneous multimedia applications like IPTV and mobile networks. However implementations of SVC specification are still limited due the augmented computational complexity. One of the most critical SVC algorithms is the motion estimation module, which allows identification of extra intra and inter-layer redundancies. The great computation demand of this module is responsible for the large SVC encoder latency time, restraining its use in real-time applications. We propose in this paper a new motion estimation approach specially designed to reduce memory accesses and data comparisons in order to increase performance. Experimental results prove significant gains in speed (up to 6 times faster) with small effects on quality (less than 0.2 dB).